Research on the Integration and Application of CNN and MRF in Geotechnics

Stratigraphic uncertainty can result in serious construction accidents or extra construction costs when the geotechnical information is limited. The Markov random field (MRF), based on the stochastic theory, provides engineers with a useful but time-consuming solution to predict the unknown stratum and relevant parameters under some priori knowledge. A new integrated model of convolutional neural network (CNN) and MRF has been proved to greatly improve the computation efficiency compared with the mentioned conventional method while keeping a relatively good accuracy, but it has not been widely used yet. In the background of rapid development of artificial intelligence (AI) and big data technology, this paper aims to introduce the above-mentioned stochastic geological modeling method to stimulate engineers to explore possible data-driven solutions to geotechnical problems. This paper firstly introduces the progress of stratigraphic uncertainty analysis, and then analyses the principles and applications in geotechnic fields of MRF and CNN. Consequently, the integrated model and a case study are introduced to prove its feasibility. Finally, the evaluation and foresight of the model is given.